Just a reiteration on prosopagnosia. Let’s assume that the findings as to the extent of face blindness pan out (I am willing to grant they found something seeing as there was an autosomal dominant pattern of inheritance in the pedigree).
1) I am skeptical that the 2% frequency in the population is due to reduced selection pressure in some populations. This behooves us to believe that we are just “catching” the German population on its way toward fixation. Additionally, this is a dominanttrait, so it seems less likely to be a loss-of-function which emerges because of relaxed selection (loss-of-function also tends toward recessiveness or additivity).
2) “Hard-wired” face recognition seems a pretty obvious trait that social animals would require to recognize conspecifics. It isn’t a lethal mutation obviously, or even highly deleterious, but as I noted, for this allele to increase in frequency and stabilize the mutation rate has to be implausibly high unless the trait is of neutral origin.
3) But neutrality, where you have mutants of equal fitness in relation to the ancestral allele, implies to me some level of polymorphism, at least in a large effective population. I suspect when you sequence this you’ll see that the 2% “face blind” carry the same allele, which likely coalesces back toward the ancestor.
4) This implies selective forces buffeting the frequency of the putative derived allele.
5) Seeing as it seems implausible that face blindness itself is the trait being selected for, it seems likely that face blindness is a byproduct of some other evolutionary dynamic. I surmised that the a dominant mutant is likely to indicate a sort of “gain of function” alteration in the protein end product of the locus, even though the phenotype of interest would imply loss of function. Well, that suggests that selection is operating on another phenotype, and the net fitness increased (now or in the past) for those with the mutant allele.
6) This fits the pattern of positively selected mutants of large effect. The Ashkenazi IQ “overclocking” mutations fall into the same category. R.A. Fisher tended to be skeptical of the power of mutations of large effect because it seemed that they were likely to “overshoot” the fitness optimum of any adaptive landscape. Genes are embedded within complex and contingent molecular networks, so large changes can result in manifold alterations in phenotype. But, consider a situation where a terrible plague sweeps over a population. If a mutant allele conferred resistance to that plague, even if it had other deleterious consequences, its net fitness would be greater than the population median. In this way rapid evolutionary changes due to selection upon alleles of large effect often result in deleterious byproducts. One assumes that over the long term “modifier” genes will emerge to mask and mitigate these deleterious byproducts, unless of course the initial selective event which flipped the fitness of the mutant allele regresses so that it is not longer advantageous. Consider malaria, it no longer confers the fitness it once did in American blacks because malaria is not a selective force in the United States. So sickle-cell anemia, the deleterious byproduct, comes to the fore as a selective force which removes the mutant generated ad hoc in response to a powerful new selective force.
The title reflects the metaphorical progress of mutations of large effect within the genome, in a given time and place they are quite likely to be “2 steps forward and 1 step behind” affairs. Populations subject to great selective stochasticity might be the perfect candidates for harboring janus faced alleles which both give and take, as ad hoc solutions come and go due to lack of stability. Do humans fit this pattern? Well, cystic fibrosis, sickle-cell anemia, eczema, etc. They might be the tip of the iceberg.